The Role of Patient–Clinician Information Engagement and Information Seeking from Nonmedical Channels in Fruit and Vegetable Intake Among Cancer Patients

Abstract

Previous research suggests positive effects of health information seeking on prevention behaviors such as diet, exercise, and fruit and vegetable consumption. The present study builds upon this research and strengthens causal claims from it by examining the lagged effect of patient–clinician information engagement on fruit and vegetable consumption as well as the indirect effect on the outcome through seeking information from nonmedical channels. The results are based on data collected from a randomly drawn sample of breast, prostate, and colorectal cancer patients from the Pennsylvania Cancer Registry who completed mail surveys in the Fall of 2006 and 2007. There was a 65% response rate for baseline subjects (resulting n = 2,013); of those, 1,293 were interviewed 1 year later, and 1,257 were available for our analyses. Results show a positive lagged main effect of patient–clinician information engagement at baseline on fruit and vegetable consumption at follow-up (B = 0.26, SE = 0.10, p = .01). The mediation analysis shows that patient–clinician information engagement leads to increased fruit and vegetable consumption among cancer patients, in part through patients' information seeking from nonmedical channels. Implications of these findings for the cancer patient population and for physicians are discussed.

Acknowledgments

The authors are grateful to Katrina Armstrong, MD, Heather Forquer, MPH, Laura Gibson, PhD, J. Sanford Schwartz, MD, and Andy S. L. Tan, MBBS, MPH, for analytic and editorial contributions.

Notes

¹The theory of planned behavior (Ajzen, 1991), the risk information seeking and processing model (Griffin, Dunwoody, & Neuwirth, 1999), the theory of motivated information management (Afifi & Weiner, 2004), the comprehensive model of information seeking (Johnson & Meischke, 1993), the health information acquisition model (Freimuth, Stein, & Kean, 1989), the integrative model of online health information seeking (Dutta-Bergman, 2006), and the extended parallel processing model (Witte, 1998).

²To note, even though we acknowledge that PRISM includes a role for affective responses to risk as triggering information seeking (Kahlor, 2010), similar to the risk information seeking and processing model (Griffin et al., 1999), this aspect is not examined in the present study.

³Past research examining mediating models of effects have traditionally drawn from the causal steps approach outlined by Baron and Kenny (1986). According to this approach, evidence for a mediating effect is supported when four conditions are fulfilled: (a) the independent variable shows a direct and significant association with the dependent variable (path c); (b) the independent variable demonstrates a significant association with the mediating variable (path a); (c) in turn, the mediating variable shows a significant association with the dependent variable (path b); and (d) the direct effect of the independent variable on the dependent variable is weakened or eliminated entirely when the effect of the mediating variable is entered into the model (path c'). Although the use of the causal steps approach is seen ubiquitously in past mediation research, MacKinnon, Fairchild, and Fritz (2007) described three important limitations to this method. Most relevant to the present study is the limitation of the first requirement in the causal steps approach stating that the independent variable must be significantly associated with the dependent variable, which is particularly problematic for studies attempting to show a lagged association between independent and dependent variables. On the contrary, MacKinnon and Fairchild (2009) noted that mediation processes do not necessarily require the presence of a direct effect between independent and dependent variables. Other limitations related to the causal steps approach also include a general weakened power to detect effects (Fritz & MacKinnon, 2007; MacKinnon, Lockwood, Hoffman, West, & Sheets, 2002), and an inability to assign a numerical value capturing the magnitude of the effect under study (MacKinnon & Fairchild, 2009).

^a Range = 0–10.

Note. All listed variables are statistically significant at p ≤ .01.

⁴Less than 1% of respondents were of Asian, American Indian, Native Hawaiian, or other descent. Among all respondents, 3% considered themselves to be Hispanic or Latino, and of these respondents, the majority also identified themselves as White. Given the distributions of participants across race/ethnicity categories, we dichotomized race/ethnicity between white and non-White categories for all analyses.

Note: Model 1 highlights the main effect of the independent factor on the outcome. Model 2 features the effect of the independent factor on the mediator. Model 3 highlights the effect of the mediator on the outcome excluding the effect of the independent factor. Model 4 adds the effect of the independent factor. *p < .05, **p < .01, ***p < .001.

⁵In a separate analysis using cross-sectional data, the association between seeking from nonmedical channels and fruit and vegetable consumption was not statistically significant. However, Maxwell and Cole (2007) observed that some mediation processes may not be captured in cross-sectional data as they require sufficient passage of time before they unfold. This is consistent with our finding a lagged association between seeking from nonmedical channels and fruit and vegetable consumption controlling for the effects of PCIE.